About the job
As NVIDIA makes inroads into the Datacenter business, our team plays a central role in getting the most out of our exponentially growing datacenter deployments as well as establishing a data-driven approach to hardware design and system software development. The role of a Deep Learning Systems Engineer would be to analyze the performance and power consumption of deep learning applications on datacenter-class hardware and significantly influence the design and optimization of datacenters.
Responsibilities
Help develop software infrastructure to characterize and analyze a broad range Deep Learning applications
Evolve cost-efficient datacenter architectures tailored to meet the needs of Large Language Models (LLMs).
Work with experts to help develop analysis and profiling tools in Python, bash and C++ to measure key performance metrics of DL workloads running on Nvidia systems.
Analyze system and software characteristics of DL applications.
Develop analysis tools and methodologies to measure key performance metrics and to estimate potential for efficiency improvement.
Qualifications
Minimum
A Bachelor’s degree in Electrical Engineering or Computer Science or equivalent experience (Masters or PhD degree preferred).
8 years or more of relevant experience.
Experience in at least one of the following:
System Software: Operating Systems (Linux), Compilers, GPU kernels (CUDA), DL Frameworks (PyTorch, TensorFlow).
Silicon Architecture and Performance Modeling/Analysis: CPU, GPU, Memory or Network Architecture
Experience programming in C/C++ and Python. Exposure to Containerization Platforms (docker) and Datacenter Workload Managers (slurm) is a plus.
A deep understanding of computer system architecture and performance analysis is essential for success in this role. Applicants should have demonstrated hands-on experience in these domains.
Demonstrated ability to work in virtual environments, and a strong drive to own tasks from beginning to end. Prior experience with such environments will make you stand out.
Preferred
Background with system software, Operating system intrinsics, GPU kernels (CUDA), or DL Frameworks (PyTorch, TensorFlow).
Experience with silicon performance monitoring or profiling tools (e.g. perf, gprof, nvidia-smi, dcgm).
In depth performance modeling experience in any one of CPU, GPU, Memory or Network Architecture
Exposure to Containerization Platforms (docker) and Datacenter Workload Managers (slurm).
Prior experience with multi-site teams or multi-functional teams.